Flowering phenology has three conventional variables: temperature, photoperiod, and vernalization. Every model of when plants flower uses some combination of these three. Some add soil moisture. Some add elevation. The models work well enough that nobody looked for a fourth variable.
Chen, Zare, and colleagues (Stanford/Chinese Academy of Sciences, PNAS 2026) found it sitting on the leaf surface every morning. Dewdrops.
Water microdroplets spontaneously generate reactive oxygen species at their air-water interface — a phenomenon Zare's group has documented for years. When dew condenses on Arabidopsis leaves, the droplets produce ROS and nitric oxide through purely abiotic interfacial chemistry. The NO transfers into the plant cells, suppresses abscisic acid biosynthesis, and accelerates flowering.
This is not a biological signal. It is a physical-chemical signal. The water droplet produces radicals because all water microdroplets produce radicals at interfaces. The plant's signaling machinery reads that chemistry as environmental information. The dew doesn't know it is signaling. The plant doesn't know it is reading dew.
Analysis of 12 million field records across the Brassicaceae family — mustards, cabbages, radishes — reveals a global correlation between dew point temperature and flowering time. The variable was there in every dataset. It was measured by every weather station. It was visible on every leaf. Nobody put it in the model because nobody thought of water as a chemical signal to plants.
The epistemological pattern: when a model with three variables fits adequately, nobody tests a fourth. The residuals get attributed to noise, local variation, genetic diversity. But adequate fit is not correct mechanism. The dew was producing radicals on every leaf surface through every dawn for a hundred million years, and the plants were responding, and the models attributed the response to temperature — which correlates with dew formation, making the omission invisible.